Public goods have a structural funding defect: the people who benefit cannot be charged, so nobody pays, so the thing that everyone relies on gets maintained by whoever happens to care. Every mechanism in this space is an attempt to route money to work that has no customer — and each one selects for a different kind of behaviour than intended.
Prospective grants select proposal-writers
The traditional model is a committee reviewing applications and funding plans. It has one dominant failure mode: it rewards the ability to write a convincing proposal, which is a skill largely uncorrelated with the ability to deliver. It also forces the funder to do the hardest thing in the process — predict which of a hundred plausible projects will matter — and to do it with less information than the applicant has.
Retroactive funding inverts the bet
Retroactive public goods funding pays for a result that already exists and can be measured. Instead of betting on a plan, you buy something that shipped.
It helps in a specific way: it selects for people who built something useful without being sure they would be paid, and it sidesteps prediction entirely. Optimism's RetroPGF is the largest live experiment, with badge-holders allocating a pool across nominated projects by category.
Two problems do not go away, they move. Measurement is now the constraint: someone still has to decide what counts as impact, and retroactive rounds have a well-documented bias toward outcomes that are legible. A widely-used library scores; a slow community-building effort that prevented a fork does not. And the cadence is hostile to maintenance, which is the least legible and most necessary work in any ecosystem — the person keeping the thing running is paid nothing retroactively for the disaster that did not happen.
Quadratic funding prices breadth
Quadratic funding changes the weighting of contributions rather than the timing. A matching pool distributes in proportion to the square of the sum of the square roots of individual contributions, so a project with many small backers receives more matching than a project with one large one.
The property is real: it prices the number of people who care rather than the amount of money they have, which is exactly the dimension a dollar-denominated market ignores. Gitcoin ran the largest public experiments with it over many rounds.
It also has two well-known attacks. Sybil attacks — one person splitting into many wallets — are the direct exploit, since the mechanism pays for apparent breadth, which is why identity verification (Passport-style scoring, or any sybil-resistant credential) is load-bearing rather than optional. And collusion defeats it more subtly: a set of projects can agree to back each other, manufacturing reciprocal matching that neither deserves.
Impact certificates make outcomes tradeable
An impact certificate turns a measured outcome into a claim someone can buy. A project delivers, a certifier verifies against a stated metric, and the certificate is minted. A funder buys it early — at a discount to the eventual payout — and is repaid if the outcome is later recognised. That is the piece that replaces "please donate" with a position someone can hold.
The hard parts are the ones you would expect. Verification is the whole game: a certificate is worth what the certifier's reputation is worth, and retroactive rounds have to resist certifiers inflating their own projects. Valuation has no natural anchor until a buyer exists, so early markets are thin and mostly subsidised. And the metric has to be checkable, which again favours legible outcomes over the maintenance work that keeps a system alive.
The domain where this is furthest along is carbon, and it is instructive that tokenising a credit does not verify it — a registry and a third-party auditor still do — but it does make retirement auditable and prevent the same credit being sold twice.
The honest synthesis
These mechanisms are complements, not competitors. Prospective grants fund things that cannot be measured but are strategically necessary. Retroactive rounds fund measurable outcomes at scale. Quadratic matching prices breadth of support. Impact certificates let individual outcomes inside a round be priced and sold.
What none of them solve is the legibility problem, and it is worth being explicit about why that matters: every mechanism here pays for what it can see, and the most important public goods are frequently the ones with no visible output. If you are funding this way, the deliberate allocation to unmeasurable maintenance is the part of the budget that proves you understood the mechanism.
References
Next, read Quadratic Funding 101. More in Governance & DAOs.